The long-term reliability of electronic components is of crucial importance in modern industry. Reliability issues arise both in terms of whether a particular component will continue to operate over time, and whether a parameter value of a component will fall within a desired range of values as the component ages. Therefore, there is an increasing demand for realistic predictions of the lifetimes of active electronic components and of the components' parameter values during those lifetimes. This same need also exists for lower level components, including conductive components such as metallization lines, wire bonds, or bumps, resistors (thick or thin film), capacitors, and dielectrics which are used for the packaging, joining and assembling of semiconductor devices. Knowledge of the physical phenomena which occur during the operation and ageing of such devices (based on drift kinetics or other models of the behavior) can be used to predict the timescale over which a device will operate within a desired range of values, or to take appropriate actions at the design level to produce a component having a desired set of characteristics.
In many cases, the reliability and lifetime of a component are estimated from accelerated ageing tests performed under well defined conditions. In a conventional ageing test designed to measure a component's resistance and predict that parameter over time, the resistance is determined for a sequence of time intervals t.sub.i at an ageing temperature T.sub.1 (where T.sub.1 &gt;room temperature). In addition, the electrical resistance is measured at room temperature in between two ageing intervals to provide a reference point. The goal is to estimate the resistance drift as a function of time based on the measured data, for times exceeding those over which the measurements are made. However, the conventional ageing method has several disadvantages:
1) The ageing treatment has to be interrupted to measure the electrical resistance at room temperature. This has the effect of restricting the number of data points obtained during a test or extending the time required to conduct the testing. The low number of data points prevents performing a realistic analysis of the ageing kinetics; PA1 2) The parameter measurement circuit has to be reconnected after each annealing (heating) step. This is time consuming and can negatively impact the resolution of the parameter measurements because of temperature induced fluctuations of the parameter value, and the need to recalibrate the measurement device; PA1 3) Room temperature fluctuations induce variations in the measured data due to the temperature coefficient of the electrical parameter being measured. For example, with a TCR (temperature coefficient of resistance) of 100 ppm (parts-per-million)/.degree.C., room temperature fluctuations of 1 .degree. C. induce a spreading of the measured results on the order of about 100 ppm. This decreases the resolution of the parameter measurements; PA1 4) The idealized temperature-time profile used to analyze the measured data is only an approximation. In reality, a certain time is needed after each resistance measurement to warm up the furnace to the ageing temperature T.sub.1, and an oversteering of the temperature can occur. Both of these effects disturb the pure isotherm character of the ageing process and introduce errors in the kinetic analysis; and PA1 5) As a consequence of the noted problems, the resolution of the measured resistance drift, dR/R.sub.0, using the conventional ageing method is rather poor, typically a few hundred ppm. A total drift which is a few orders of magnitude larger than the measurement resolution is generally required to properly perform an ageing analysis based on kinetics theory. Unfortunately, such a drift can only be achieved after several thousand hours of ageing using good quality materials. This makes such measurements impractical for many applications.
Because of the time-consuming character of placing a component having an ageing parameter in an environment with a raised temperature, subsequently measuring the changed parameter, and repeating this cycle until sufficient measurements are made to permit a reliable prediction of the parameter value over time, this method is particularly disadvantageous for systems where a small change of the parameters occurs. With the described method a typical value for the measurement error is 100 ppm. In such a situation an ageing to 10,000 ppm is necessary in order to predict the parameter value as a function of time based on extrapolations of the change in the parameter during the test period. Measuring wires, such as current and voltage wires have to be connected and disconnected again each time, while some time is necessary in each case to reach a desired degree of stability and temperature of the oven. The instability of the oven temperature when switching measurement and heating equipment on and off also contributes to inaccuracies in the measurements.
In addition, measurements may be carried out under various conditions, such as in a helium atmosphere, in air, in polluted air for sensors used in determining the degree of air pollution, in an 0.sub.2 atmosphere, or in an atmosphere of moistened air. In such situations it is important that the desired conditions are maintained within an acceptable tolerance range during the course of the measurements. This can become difficult to achieve if the equipment containing the component is regularly opened up in order to make measurements of an ageing parameter.
Component reliability is also sometimes examined by stressing a device in order to determine if it will operate properly. This is done to identify components which fail to operate properly so that they can be segregated from those that successfully pass the stress test. This type of testing, often termed "burn-in testing" is actually a form of quality control rather than a determination of the value of an ageing parameter. Such testing methods can be used to determine if a component will operate under the conditions of the test, but are not suited to determining the value of a component's electromagnetic parameter over the lifetime of the component.
U.S. Pat. No. 4,374,317, issued Feb. 15, 1983 to Bradshaw, discloses an improved means of connecting a component board located within an oven to control circuitry as part of "burning-in" the component board. Bradshaw's burning-in method includes subjecting the board to a high temperature environment, operating the component(s) contained on the board at that high temperature, and detecting failures of the components. Bradshaw does not teach conducting in-situ measurements of the components. Furthermore, Bradshaw is concerned with determining how well a component or components operate when stressed, so that components which fail to operate properly can be segregated from those that pass the burn-in test. Bradshaw is not concerned with predicting how a component will function over its lifetime by determining how its parameter value changes.
U.S. Pat. No. 3,842,346, issued Oct. 15, 1974 to Bobbitt, discloses an apparatus for and method of testing integrated circuits during a change in temperature to determine if the circuits maintain their electrical continuity. This is done for the purpose of identifying any intermittent circuit connections which may be present. Thus, Bobbitt is also directed to quality control of the tested circuits by assuring that the circuits will operate under the thermal conditions to which they are exposed.
What is desired is an apparatus and method for making reliable high resolution measurements of a component's electromagnetic ageing parameter within a shorter time period than required by current techniques. It is also desired to have a method for using these measurements to predict a component's electromagnetic parameter values as the component ages.